A very high percentage of the cost of an application is its design, development and maintenance. By comparison, the cost of running the application is usually so much smaller it is not worth worrying about. So when you think about efficiency, you should think about your efficiency and those who have to maintain your application after you (assuming your application will be useful ;)
Say you have application where performance is an issue. A common mistake is to just guess where the performance issue are and "optimise" those. The problem with this approach is that it usually
- adds complexity
- reduces maintainability.
- misses the biggest performance bottlenecks
- can make little or no difference or can even make performance worse. (The last time some one tried to optimise some code I had written it was 50% slower ;)
What you need to do is to perform some realistic tests and measure the performance bottlenecks which exceed your specific measurable business requirements A vague go-as-fast-as-possible is not specific or measurable, and usually non-sense in my experience.
Once you have made the changes to the key bottlenecks you need to re-test the system to ensure it really helped.